After last week’s post, one or two people pointed out to me that it’s okay to use xG in analyzing single matches, as it at least gives us a more accurate idea of which team created the better chances on the day. The idea here is that while goals are a matter of luck, the ability to create chances is not.
While I think this contains a kernel of truth, it is misleading. A better way to rephrase this might be, “goals are more subject to luck than the ability to create chances.”
While this distinction may come off as pedantic, it’s essential, and one of the main reasons why involves the use of xG in single games.
The rationale behind adding up xG totals for single matches, on the face of it, makes a lot of sense. After all, if the ability to create good chances is more consistent and predictable than the ability to score goals, then understanding the xG total might give us a more accurate picture of which team had the better performance. This omits the fact that xG’s predictive power only stabilizes (last time I checked, and this is off the top of my head) at around 5-10 matches.
Defenders of single match xG totalling often defend this by arguing that while a single game may not tell us that how one team will perform over the next five or ten matches, it at least shows us which side had the better performance on the day.
But it doesn’t, for the same reason that goals scored doesn’t necessarily tell us which team was dominant—it is, at 90 minutes, still prone to a lot of random variation, or statistical noise. I could, of course, use this as a platform to theorize about the many possible, practical reasons as to why this is. For example, perhaps the better team had an off-day, or perhaps one side worked the ball far more expertly in the final third but was more cautious about shooting and therefore registered a smaller xG.
But the reasons are moot. If xG has little significant predictive power after a single game, it cannot be used to make qualitative statements about team performance, full stop. This is why I think we ought to take seriously David Sumpter’s call to omit the use of xG completely in analyzing 1-2 matches.
So then how do we qualify an excellent individual performance on the day? Use your eyes, and your best judgment. Just like a decent football journalist might.